1 files changed, 174 insertions, 218 deletions

diff --git a/drivers/block/ll_rw_blk.c b/drivers/block/ll_rw_blk.c
index f20eba22b14b..692a5fced76e 100644
--- a/drivers/block/ll_rw_blk.c
+++ b/drivers/block/ll_rw_blk.c
@@ -28,6 +28,7 @@
 #include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
+#include <linux/blkdev.h>
 
 /*
  * for max sense size
@@ -36,6 +37,7 @@
 
 static void blk_unplug_work(void *data);
 static void blk_unplug_timeout(unsigned long data);
+static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io);
 
 /*
  * For the allocated request tables
@@ -274,6 +276,7 @@ static inline void rq_init(request_queue_t *q, struct request *rq)
 	rq->errors = 0;
 	rq->rq_status = RQ_ACTIVE;
 	rq->bio = rq->biotail = NULL;
+	rq->ioprio = 0;
 	rq->buffer = NULL;
 	rq->ref_count = 1;
 	rq->q = q;
@@ -716,7 +719,7 @@ struct request *blk_queue_find_tag(request_queue_t *q, int tag)
 {
 	struct blk_queue_tag *bqt = q->queue_tags;
 
-	if (unlikely(bqt == NULL || tag >= bqt->real_max_depth))
+	if (unlikely(bqt == NULL || tag >= bqt->max_depth))
 		return NULL;
 
 	return bqt->tag_index[tag];
@@ -774,9 +777,9 @@ EXPORT_SYMBOL(blk_queue_free_tags);
 static int
 init_tag_map(request_queue_t *q, struct blk_queue_tag *tags, int depth)
 {
-	int bits, i;
 	struct request **tag_index;
 	unsigned long *tag_map;
+	int nr_ulongs;
 
 	if (depth > q->nr_requests * 2) {
 		depth = q->nr_requests * 2;
@@ -788,24 +791,17 @@ init_tag_map(request_queue_t *q, struct blk_queue_tag *tags, int depth)
 	if (!tag_index)
 		goto fail;
 
-	bits = (depth / BLK_TAGS_PER_LONG) + 1;
-	tag_map = kmalloc(bits * sizeof(unsigned long), GFP_ATOMIC);
+	nr_ulongs = ALIGN(depth, BITS_PER_LONG) / BITS_PER_LONG;
+	tag_map = kmalloc(nr_ulongs * sizeof(unsigned long), GFP_ATOMIC);
 	if (!tag_map)
 		goto fail;
 
 	memset(tag_index, 0, depth * sizeof(struct request *));
-	memset(tag_map, 0, bits * sizeof(unsigned long));
+	memset(tag_map, 0, nr_ulongs * sizeof(unsigned long));
 	tags->max_depth = depth;
-	tags->real_max_depth = bits * BITS_PER_LONG;
 	tags->tag_index = tag_index;
 	tags->tag_map = tag_map;
 
-	/*
-	 * set the upper bits if the depth isn't a multiple of the word size
-	 */
-	for (i = depth; i < bits * BLK_TAGS_PER_LONG; i++)
-		__set_bit(i, tag_map);
-
 	return 0;
 fail:
 	kfree(tag_index);
@@ -870,32 +866,24 @@ int blk_queue_resize_tags(request_queue_t *q, int new_depth)
 	struct blk_queue_tag *bqt = q->queue_tags;
 	struct request **tag_index;
 	unsigned long *tag_map;
-	int bits, max_depth;
+	int max_depth, nr_ulongs;
 
 	if (!bqt)
 		return -ENXIO;
 
 	/*
-	 * don't bother sizing down
-	 */
-	if (new_depth <= bqt->real_max_depth) {
-		bqt->max_depth = new_depth;
-		return 0;
-	}
-
-	/*
 	 * save the old state info, so we can copy it back
 	 */
 	tag_index = bqt->tag_index;
 	tag_map = bqt->tag_map;
-	max_depth = bqt->real_max_depth;
+	max_depth = bqt->max_depth;
 
 	if (init_tag_map(q, bqt, new_depth))
 		return -ENOMEM;
 
 	memcpy(bqt->tag_index, tag_index, max_depth * sizeof(struct request *));
-	bits = max_depth / BLK_TAGS_PER_LONG;
-	memcpy(bqt->tag_map, tag_map, bits * sizeof(unsigned long));
+	nr_ulongs = ALIGN(max_depth, BITS_PER_LONG) / BITS_PER_LONG;
+	memcpy(bqt->tag_map, tag_map, nr_ulongs * sizeof(unsigned long));
 
 	kfree(tag_index);
 	kfree(tag_map);
@@ -925,11 +913,16 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
 
 	BUG_ON(tag == -1);
 
-	if (unlikely(tag >= bqt->real_max_depth))
+	if (unlikely(tag >= bqt->max_depth))
+		/*
+		 * This can happen after tag depth has been reduced.
+		 * FIXME: how about a warning or info message here?
+		 */
 		return;
 
 	if (unlikely(!__test_and_clear_bit(tag, bqt->tag_map))) {
-		printk("attempt to clear non-busy tag (%d)\n", tag);
+		printk(KERN_ERR "%s: attempt to clear non-busy tag (%d)\n",
+		       __FUNCTION__, tag);
 		return;
 	}
 
@@ -938,7 +931,8 @@ void blk_queue_end_tag(request_queue_t *q, struct request *rq)
 	rq->tag = -1;
 
 	if (unlikely(bqt->tag_index[tag] == NULL))
-		printk("tag %d is missing\n", tag);
+		printk(KERN_ERR "%s: tag %d is missing\n",
+		       __FUNCTION__, tag);
 
 	bqt->tag_index[tag] = NULL;
 	bqt->busy--;
@@ -967,24 +961,20 @@ EXPORT_SYMBOL(blk_queue_end_tag);
 int blk_queue_start_tag(request_queue_t *q, struct request *rq)
 {
 	struct blk_queue_tag *bqt = q->queue_tags;
-	unsigned long *map = bqt->tag_map;
-	int tag = 0;
+	int tag;
 
 	if (unlikely((rq->flags & REQ_QUEUED))) {
 		printk(KERN_ERR 
-		       "request %p for device [%s] already tagged %d",
-		       rq, rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->tag);
+		       "%s: request %p for device [%s] already tagged %d",
+		       __FUNCTION__, rq,
+		       rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->tag);
 		BUG();
 	}
 
-	for (map = bqt->tag_map; *map == -1UL; map++) {
-		tag += BLK_TAGS_PER_LONG;
-
-		if (tag >= bqt->max_depth)
-			return 1;
-	}
+	tag = find_first_zero_bit(bqt->tag_map, bqt->max_depth);
+	if (tag >= bqt->max_depth)
+		return 1;
 
-	tag += ffz(*map);
 	__set_bit(tag, bqt->tag_map);
 
 	rq->flags |= REQ_QUEUED;
@@ -1020,7 +1010,8 @@ void blk_queue_invalidate_tags(request_queue_t *q)
 		rq = list_entry_rq(tmp);
 
 		if (rq->tag == -1) {
-			printk("bad tag found on list\n");
+			printk(KERN_ERR
+			       "%s: bad tag found on list\n", __FUNCTION__);
 			list_del_init(&rq->queuelist);
 			rq->flags &= ~REQ_QUEUED;
 		} else
@@ -1148,7 +1139,7 @@ new_hw_segment:
 }
 
 
-int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
 				   struct bio *nxt)
 {
 	if (!(q->queue_flags & (1 << QUEUE_FLAG_CLUSTER)))
@@ -1169,9 +1160,7 @@ int blk_phys_contig_segment(request_queue_t *q, struct bio *bio,
 	return 0;
 }
 
-EXPORT_SYMBOL(blk_phys_contig_segment);
-
-int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
+static int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
 				 struct bio *nxt)
 {
 	if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
@@ -1187,8 +1176,6 @@ int blk_hw_contig_segment(request_queue_t *q, struct bio *bio,
 	return 1;
 }
 
-EXPORT_SYMBOL(blk_hw_contig_segment);
-
 /*
  * map a request to scatterlist, return number of sg entries setup. Caller
  * must make sure sg can hold rq->nr_phys_segments entries
@@ -1358,8 +1345,8 @@ static int ll_front_merge_fn(request_queue_t *q, struct request *req,
 static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
 				struct request *next)
 {
-	int total_phys_segments = req->nr_phys_segments +next->nr_phys_segments;
-	int total_hw_segments = req->nr_hw_segments + next->nr_hw_segments;
+	int total_phys_segments;
+	int total_hw_segments;
 
 	/*
 	 * First check if the either of the requests are re-queued
@@ -1369,7 +1356,7 @@ static int ll_merge_requests_fn(request_queue_t *q, struct request *req,
 		return 0;
 
 	/*
-	 * Will it become to large?
+	 * Will it become too large?
 	 */
 	if ((req->nr_sectors + next->nr_sectors) > q->max_sectors)
 		return 0;
@@ -1450,17 +1437,13 @@ EXPORT_SYMBOL(blk_remove_plug);
  */
 void __generic_unplug_device(request_queue_t *q)
 {
-	if (test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags))
+	if (unlikely(test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)))
 		return;
 
 	if (!blk_remove_plug(q))
 		return;
 
-	/*
-	 * was plugged, fire request_fn if queue has stuff to do
-	 */
-	if (elv_next_request(q))
-		q->request_fn(q);
+	q->request_fn(q);
 }
 EXPORT_SYMBOL(__generic_unplug_device);
 
@@ -1645,7 +1628,8 @@ static int blk_init_free_list(request_queue_t *q)
 	init_waitqueue_head(&rl->wait[WRITE]);
 	init_waitqueue_head(&rl->drain);
 
-	rl->rq_pool = mempool_create(BLKDEV_MIN_RQ, mempool_alloc_slab, mempool_free_slab, request_cachep);
+	rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ, mempool_alloc_slab,
+				mempool_free_slab, request_cachep, q->node);
 
 	if (!rl->rq_pool)
 		return -ENOMEM;
@@ -1657,8 +1641,15 @@ static int __make_request(request_queue_t *, struct bio *);
 
 request_queue_t *blk_alloc_queue(int gfp_mask)
 {
-	request_queue_t *q = kmem_cache_alloc(requestq_cachep, gfp_mask);
+	return blk_alloc_queue_node(gfp_mask, -1);
+}
+EXPORT_SYMBOL(blk_alloc_queue);
 
+request_queue_t *blk_alloc_queue_node(int gfp_mask, int node_id)
+{
+	request_queue_t *q;
+
+	q = kmem_cache_alloc_node(requestq_cachep, gfp_mask, node_id);
 	if (!q)
 		return NULL;
 
@@ -1671,8 +1662,7 @@ request_queue_t *blk_alloc_queue(int gfp_mask)
 
 	return q;
 }
-
-EXPORT_SYMBOL(blk_alloc_queue);
+EXPORT_SYMBOL(blk_alloc_queue_node);
 
 /**
  * blk_init_queue  - prepare a request queue for use with a block device
@@ -1705,13 +1695,22 @@ EXPORT_SYMBOL(blk_alloc_queue);
  *    blk_init_queue() must be paired with a blk_cleanup_queue() call
  *    when the block device is deactivated (such as at module unload).
  **/
+
 request_queue_t *blk_init_queue(request_fn_proc *rfn, spinlock_t *lock)
 {
-	request_queue_t *q = blk_alloc_queue(GFP_KERNEL);
+	return blk_init_queue_node(rfn, lock, -1);
+}
+EXPORT_SYMBOL(blk_init_queue);
+
+request_queue_t *
+blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
+{
+	request_queue_t *q = blk_alloc_queue_node(GFP_KERNEL, node_id);
 
 	if (!q)
 		return NULL;
 
+	q->node = node_id;
 	if (blk_init_free_list(q))
 		goto out_init;
 
@@ -1754,12 +1753,11 @@ out_init:
 	kmem_cache_free(requestq_cachep, q);
 	return NULL;
 }
-
-EXPORT_SYMBOL(blk_init_queue);
+EXPORT_SYMBOL(blk_init_queue_node);
 
 int blk_get_queue(request_queue_t *q)
 {
-	if (!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)) {
+	if (likely(!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) {
 		atomic_inc(&q->refcnt);
 		return 0;
 	}
@@ -1775,8 +1773,8 @@ static inline void blk_free_request(request_queue_t *q, struct request *rq)
 	mempool_free(rq, q->rq.rq_pool);
 }
 
-static inline struct request *blk_alloc_request(request_queue_t *q, int rw,
-						int gfp_mask)
+static inline struct request *
+blk_alloc_request(request_queue_t *q, int rw, struct bio *bio, int gfp_mask)
 {
 	struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask);
 
@@ -1789,7 +1787,7 @@ static inline struct request *blk_alloc_request(request_queue_t *q, int rw,
 	 */
 	rq->flags = rw;
 
-	if (!elv_set_request(q, rq, gfp_mask))
+	if (!elv_set_request(q, rq, bio, gfp_mask))
 		return rq;
 
 	mempool_free(rq, q->rq.rq_pool);
@@ -1821,7 +1819,7 @@ static inline int ioc_batching(request_queue_t *q, struct io_context *ioc)
  * is the behaviour we want though - once it gets a wakeup it should be given
  * a nice run.
  */
-void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
+static void ioc_set_batching(request_queue_t *q, struct io_context *ioc)
 {
 	if (!ioc || ioc_batching(q, ioc))
 		return;
@@ -1838,7 +1836,6 @@ static void __freed_request(request_queue_t *q, int rw)
 		clear_queue_congested(q, rw);
 
 	if (rl->count[rw] + 1 <= q->nr_requests) {
-		smp_mb();
 		if (waitqueue_active(&rl->wait[rw]))
 			wake_up(&rl->wait[rw]);
 
@@ -1870,18 +1867,20 @@ static void freed_request(request_queue_t *q, int rw)
 
 #define blkdev_free_rq(list) list_entry((list)->next, struct request, queuelist)
 /*
- * Get a free request, queue_lock must not be held
+ * Get a free request, queue_lock must be held.
+ * Returns NULL on failure, with queue_lock held.
+ * Returns !NULL on success, with queue_lock *not held*.
  */
-static struct request *get_request(request_queue_t *q, int rw, int gfp_mask)
+static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
+				   int gfp_mask)
 {
 	struct request *rq = NULL;
 	struct request_list *rl = &q->rq;
-	struct io_context *ioc = get_io_context(gfp_mask);
+	struct io_context *ioc = current_io_context(GFP_ATOMIC);
 
 	if (unlikely(test_bit(QUEUE_FLAG_DRAIN, &q->queue_flags)))
 		goto out;
 
-	spin_lock_irq(q->queue_lock);
 	if (rl->count[rw]+1 >= q->nr_requests) {
 		/*
 		 * The queue will fill after this allocation, so set it as
@@ -1895,7 +1894,7 @@ static struct request *get_request(request_queue_t *q, int rw, int gfp_mask)
 		}
 	}
 
-	switch (elv_may_queue(q, rw)) {
+	switch (elv_may_queue(q, rw, bio)) {
 		case ELV_MQUEUE_NO:
 			goto rq_starved;
 		case ELV_MQUEUE_MAY:
@@ -1909,18 +1908,25 @@ static struct request *get_request(request_queue_t *q, int rw, int gfp_mask)
 		 * The queue is full and the allocating process is not a
 		 * "batcher", and not exempted by the IO scheduler
 		 */
-		spin_unlock_irq(q->queue_lock);
 		goto out;
 	}
 
 get_rq:
+	/*
+	 * Only allow batching queuers to allocate up to 50% over the defined
+	 * limit of requests, otherwise we could have thousands of requests
+	 * allocated with any setting of ->nr_requests
+	 */
+	if (rl->count[rw] >= (3 * q->nr_requests / 2))
+		goto out;
+
 	rl->count[rw]++;
 	rl->starved[rw] = 0;
 	if (rl->count[rw] >= queue_congestion_on_threshold(q))
 		set_queue_congested(q, rw);
 	spin_unlock_irq(q->queue_lock);
 
-	rq = blk_alloc_request(q, rw, gfp_mask);
+	rq = blk_alloc_request(q, rw, bio, gfp_mask);
 	if (!rq) {
 		/*
 		 * Allocation failed presumably due to memory. Undo anything
@@ -1943,7 +1949,6 @@ rq_starved:
 		if (unlikely(rl->count[rw] == 0))
 			rl->starved[rw] = 1;
 
-		spin_unlock_irq(q->queue_lock);
 		goto out;
 	}
 
@@ -1953,31 +1958,35 @@ rq_starved:
 	rq_init(q, rq);
 	rq->rl = rl;
 out:
-	put_io_context(ioc);
 	return rq;
 }
 
 /*
  * No available requests for this queue, unplug the device and wait for some
  * requests to become available.
+ *
+ * Called with q->queue_lock held, and returns with it unlocked.
  */
-static struct request *get_request_wait(request_queue_t *q, int rw)
+static struct request *get_request_wait(request_queue_t *q, int rw,
+					struct bio *bio)
 {
-	DEFINE_WAIT(wait);
 	struct request *rq;
 
-	generic_unplug_device(q);
-	do {
+	rq = get_request(q, rw, bio, GFP_NOIO);
+	while (!rq) {
+		DEFINE_WAIT(wait);
 		struct request_list *rl = &q->rq;
 
 		prepare_to_wait_exclusive(&rl->wait[rw], &wait,
 				TASK_UNINTERRUPTIBLE);
 
-		rq = get_request(q, rw, GFP_NOIO);
+		rq = get_request(q, rw, bio, GFP_NOIO);
 
 		if (!rq) {
 			struct io_context *ioc;
 
+			__generic_unplug_device(q);
+			spin_unlock_irq(q->queue_lock);
 			io_schedule();
 
 			/*
@@ -1986,12 +1995,13 @@ static struct request *get_request_wait(request_queue_t *q, int rw)
 			 * up to a big batch of them for a small period time.
 			 * See ioc_batching, ioc_set_batching
 			 */
-			ioc = get_io_context(GFP_NOIO);
+			ioc = current_io_context(GFP_NOIO);
 			ioc_set_batching(q, ioc);
-			put_io_context(ioc);
+
+			spin_lock_irq(q->queue_lock);
 		}
 		finish_wait(&rl->wait[rw], &wait);
-	} while (!rq);
+	}
 
 	return rq;
 }
@@ -2002,14 +2012,18 @@ struct request *blk_get_request(request_queue_t *q, int rw, int gfp_mask)
 
 	BUG_ON(rw != READ && rw != WRITE);
 
-	if (gfp_mask & __GFP_WAIT)
-		rq = get_request_wait(q, rw);
-	else
-		rq = get_request(q, rw, gfp_mask);
+	spin_lock_irq(q->queue_lock);
+	if (gfp_mask & __GFP_WAIT) {
+		rq = get_request_wait(q, rw, NULL);
+	} else {
+		rq = get_request(q, rw, NULL, gfp_mask);
+		if (!rq)
+			spin_unlock_irq(q->queue_lock);
+	}
+	/* q->queue_lock is unlocked at this point */
 
 	return rq;
 }
-
 EXPORT_SYMBOL(blk_get_request);
 
 /**
@@ -2251,45 +2265,7 @@ int blkdev_issue_flush(struct block_device *bdev, sector_t *error_sector)
 
 EXPORT_SYMBOL(blkdev_issue_flush);
 
-/**
- * blkdev_scsi_issue_flush_fn - issue flush for SCSI devices
- * @q:		device queue
- * @disk:	gendisk
- * @error_sector:	error offset
- *
- * Description:
- *    Devices understanding the SCSI command set, can use this function as
- *    a helper for issuing a cache flush. Note: driver is required to store
- *    the error offset (in case of error flushing) in ->sector of struct
- *    request.
- */
-int blkdev_scsi_issue_flush_fn(request_queue_t *q, struct gendisk *disk,
-			       sector_t *error_sector)
-{
-	struct request *rq = blk_get_request(q, WRITE, __GFP_WAIT);
-	int ret;
-
-	rq->flags |= REQ_BLOCK_PC | REQ_SOFTBARRIER;
-	rq->sector = 0;
-	memset(rq->cmd, 0, sizeof(rq->cmd));
-	rq->cmd[0] = 0x35;
-	rq->cmd_len = 12;
-	rq->data = NULL;
-	rq->data_len = 0;
-	rq->timeout = 60 * HZ;
-
-	ret = blk_execute_rq(q, disk, rq);
-
-	if (ret && error_sector)
-		*error_sector = rq->sector;
-
-	blk_put_request(rq);
-	return ret;
-}
-
-EXPORT_SYMBOL(blkdev_scsi_issue_flush_fn);
-
-void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
+static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io)
 {
 	int rw = rq_data_dir(rq);
 
@@ -2371,7 +2347,6 @@ static void __blk_put_request(request_queue_t *q, struct request *req)
 		return;
 
 	req->rq_status = RQ_INACTIVE;
-	req->q = NULL;
 	req->rl = NULL;
 
 	/*
@@ -2500,6 +2475,8 @@ static int attempt_merge(request_queue_t *q, struct request *req,
 		req->rq_disk->in_flight--;
 	}
 
+	req->ioprio = ioprio_best(req->ioprio, next->ioprio);
+
 	__blk_put_request(q, next);
 	return 1;
 }
@@ -2548,25 +2525,17 @@ void blk_attempt_remerge(request_queue_t *q, struct request *rq)
 
 EXPORT_SYMBOL(blk_attempt_remerge);
 
-/*
- * Non-locking blk_attempt_remerge variant.
- */
-void __blk_attempt_remerge(request_queue_t *q, struct request *rq)
-{
-	attempt_back_merge(q, rq);
-}
-
-EXPORT_SYMBOL(__blk_attempt_remerge);
-
 static int __make_request(request_queue_t *q, struct bio *bio)
 {
-	struct request *req, *freereq = NULL;
+	struct request *req;
 	int el_ret, rw, nr_sectors, cur_nr_sectors, barrier, err, sync;
+	unsigned short prio;
 	sector_t sector;
 
 	sector = bio->bi_sector;
 	nr_sectors = bio_sectors(bio);
 	cur_nr_sectors = bio_cur_sectors(bio);
+	prio = bio_prio(bio);
 
 	rw = bio_data_dir(bio);
 	sync = bio_sync(bio);
@@ -2581,19 +2550,14 @@ static int __make_request(request_queue_t *q, struct bio *bio)
 	spin_lock_prefetch(q->queue_lock);
 
 	barrier = bio_barrier(bio);
-	if (barrier && (q->ordered == QUEUE_ORDERED_NONE)) {
+	if (unlikely(barrier) && (q->ordered == QUEUE_ORDERED_NONE)) {
 		err = -EOPNOTSUPP;
 		goto end_io;
 	}
 
-again:
 	spin_lock_irq(q->queue_lock);
 
-	if (elv_queue_empty(q)) {
-		blk_plug_device(q);
-		goto get_rq;
-	}
-	if (barrier)
+	if (unlikely(barrier) || elv_queue_empty(q))
 		goto get_rq;
 
 	el_ret = elv_merge(q, &req, bio);
@@ -2607,6 +2571,7 @@ again:
 			req->biotail->bi_next = bio;
 			req->biotail = bio;
 			req->nr_sectors = req->hard_nr_sectors += nr_sectors;
+			req->ioprio = ioprio_best(req->ioprio, prio);
 			drive_stat_acct(req, nr_sectors, 0);
 			if (!attempt_back_merge(q, req))
 				elv_merged_request(q, req);
@@ -2631,45 +2596,30 @@ again:
 			req->hard_cur_sectors = cur_nr_sectors;
 			req->sector = req->hard_sector = sector;
 			req->nr_sectors = req->hard_nr_sectors += nr_sectors;
+			req->ioprio = ioprio_best(req->ioprio, prio);
 			drive_stat_acct(req, nr_sectors, 0);
 			if (!attempt_front_merge(q, req))
 				elv_merged_request(q, req);
 			goto out;
 
-		/*
-		 * elevator says don't/can't merge. get new request
-		 */
-		case ELEVATOR_NO_MERGE:
-			break;
-
+		/* ELV_NO_MERGE: elevator says don't/can't merge. */
 		default:
-			printk("elevator returned crap (%d)\n", el_ret);
-			BUG();
+			;
 	}
 
+get_rq:
 	/*
-	 * Grab a free request from the freelist - if that is empty, check
-	 * if we are doing read ahead and abort instead of blocking for
-	 * a free slot.
+	 * Grab a free request. This is might sleep but can not fail.
+	 * Returns with the queue unlocked.
+	 */
+	req = get_request_wait(q, rw, bio);
+
+	/*
+	 * After dropping the lock and possibly sleeping here, our request
+	 * may now be mergeable after it had proven unmergeable (above).
+	 * We don't worry about that case for efficiency. It won't happen
+	 * often, and the elevators are able to handle it.
 	 */
-get_rq:
-	if (freereq) {
-		req = freereq;
-		freereq = NULL;
-	} else {
-		spin_unlock_irq(q->queue_lock);
-		if ((freereq = get_request(q, rw, GFP_ATOMIC)) == NULL) {
-			/*
-			 * READA bit set
-			 */
-			err = -EWOULDBLOCK;
-			if (bio_rw_ahead(bio))
-				goto end_io;
-	
-			freereq = get_request_wait(q, rw);
-		}
-		goto again;
-	}
 
 	req->flags |= REQ_CMD;
 
@@ -2682,7 +2632,7 @@ get_rq:
 	/*
 	 * REQ_BARRIER implies no merging, but lets make it explicit
 	 */
-	if (barrier)
+	if (unlikely(barrier))
 		req->flags |= (REQ_HARDBARRIER | REQ_NOMERGE);
 
 	req->errors = 0;
@@ -2694,13 +2644,15 @@ get_rq:
 	req->buffer = bio_data(bio);	/* see ->buffer comment above */
 	req->waiting = NULL;
 	req->bio = req->biotail = bio;
+	req->ioprio = prio;
 	req->rq_disk = bio->bi_bdev->bd_disk;
 	req->start_time = jiffies;
 
+	spin_lock_irq(q->queue_lock);
+	if (elv_queue_empty(q))
+		blk_plug_device(q);
 	add_request(q, req);
 out:
-	if (freereq)
-		__blk_put_request(q, freereq);
 	if (sync)
 		__generic_unplug_device(q);
 
@@ -2722,7 +2674,7 @@ static inline void blk_partition_remap(struct bio *bio)
 	if (bdev != bdev->bd_contains) {
 		struct hd_struct *p = bdev->bd_part;
 
-		switch (bio->bi_rw) {
+		switch (bio_data_dir(bio)) {
 		case READ:
 			p->read_sectors += bio_sectors(bio);
 			p->reads++;
@@ -2741,6 +2693,7 @@ void blk_finish_queue_drain(request_queue_t *q)
 {
 	struct request_list *rl = &q->rq;
 	struct request *rq;
+	int requeued = 0;
 
 	spin_lock_irq(q->queue_lock);
 	clear_bit(QUEUE_FLAG_DRAIN, &q->queue_flags);
@@ -2749,9 +2702,13 @@ void blk_finish_queue_drain(request_queue_t *q)
 		rq = list_entry_rq(q->drain_list.next);
 
 		list_del_init(&rq->queuelist);
-		__elv_add_request(q, rq, ELEVATOR_INSERT_BACK, 1);
+		elv_requeue_request(q, rq);
+		requeued++;
 	}
 
+	if (requeued)
+		q->request_fn(q);
+
 	spin_unlock_irq(q->queue_lock);
 
 	wake_up(&rl->wait[0]);
@@ -2806,7 +2763,7 @@ static inline void block_wait_queue_running(request_queue_t *q)
 {
 	DEFINE_WAIT(wait);
 
-	while (test_bit(QUEUE_FLAG_DRAIN, &q->queue_flags)) {
+	while (unlikely(test_bit(QUEUE_FLAG_DRAIN, &q->queue_flags))) {
 		struct request_list *rl = &q->rq;
 
 		prepare_to_wait_exclusive(&rl->drain, &wait,
@@ -2915,7 +2872,7 @@ end_io:
 			goto end_io;
 		}
 
-		if (test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))
+		if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)))
 			goto end_io;
 
 		block_wait_queue_running(q);
@@ -2948,7 +2905,7 @@ void submit_bio(int rw, struct bio *bio)
 
 	BIO_BUG_ON(!bio->bi_size);
 	BIO_BUG_ON(!bio->bi_io_vec);
-	bio->bi_rw = rw;
+	bio->bi_rw |= rw;
 	if (rw & WRITE)
 		mod_page_state(pgpgout, count);
 	else
@@ -2968,7 +2925,7 @@ void submit_bio(int rw, struct bio *bio)
 
 EXPORT_SYMBOL(submit_bio);
 
-void blk_recalc_rq_segments(struct request *rq)
+static void blk_recalc_rq_segments(struct request *rq)
 {
 	struct bio *bio, *prevbio = NULL;
 	int nr_phys_segs, nr_hw_segs;
@@ -3010,7 +2967,7 @@ void blk_recalc_rq_segments(struct request *rq)
 	rq->nr_hw_segments = nr_hw_segs;
 }
 
-void blk_recalc_rq_sectors(struct request *rq, int nsect)
+static void blk_recalc_rq_sectors(struct request *rq, int nsect)
 {
 	if (blk_fs_request(rq)) {
 		rq->hard_sector += nsect;
@@ -3305,8 +3262,11 @@ void exit_io_context(void)
 	struct io_context *ioc;
 
 	local_irq_save(flags);
+	task_lock(current);
 	ioc = current->io_context;
 	current->io_context = NULL;
+	ioc->task = NULL;
+	task_unlock(current);
 	local_irq_restore(flags);
 
 	if (ioc->aic && ioc->aic->exit)
@@ -3319,53 +3279,49 @@ void exit_io_context(void)
 
 /*
  * If the current task has no IO context then create one and initialise it.
- * If it does have a context, take a ref on it.
+ * Otherwise, return its existing IO context.
  *
- * This is always called in the context of the task which submitted the I/O.
- * But weird things happen, so we disable local interrupts to ensure exclusive
- * access to *current.
+ * This returned IO context doesn't have a specifically elevated refcount,
+ * but since the current task itself holds a reference, the context can be
+ * used in general code, so long as it stays within `current` context.
  */
-struct io_context *get_io_context(int gfp_flags)
+struct io_context *current_io_context(int gfp_flags)
 {
 	struct task_struct *tsk = current;
-	unsigned long flags;
 	struct io_context *ret;
 
-	local_irq_save(flags);
 	ret = tsk->io_context;
-	if (ret)
-		goto out;
-
-	local_irq_restore(flags);
+	if (likely(ret))
+		return ret;
 
 	ret = kmem_cache_alloc(iocontext_cachep, gfp_flags);
 	if (ret) {
 		atomic_set(&ret->refcount, 1);
-		ret->pid = tsk->pid;
+		ret->task = current;
+		ret->set_ioprio = NULL;
 		ret->last_waited = jiffies; /* doesn't matter... */
 		ret->nr_batch_requests = 0; /* because this is 0 */
 		ret->aic = NULL;
 		ret->cic = NULL;
-		spin_lock_init(&ret->lock);
-
-		local_irq_save(flags);
+		tsk->io_context = ret;
+	}
 
-		/*
-		 * very unlikely, someone raced with us in setting up the task
-		 * io context. free new context and just grab a reference.
-		 */
-		if (!tsk->io_context)
-			tsk->io_context = ret;
-		else {
-			kmem_cache_free(iocontext_cachep, ret);
-			ret = tsk->io_context;
-		}
+	return ret;
+}
+EXPORT_SYMBOL(current_io_context);
 
-out:
+/*
+ * If the current task has no IO context then create one and initialise it.
+ * If it does have a context, take a ref on it.
+ *
+ * This is always called in the context of the task which submitted the I/O.
+ */
+struct io_context *get_io_context(int gfp_flags)
+{
+	struct io_context *ret;
+	ret = current_io_context(gfp_flags);
+	if (likely(ret))
 		atomic_inc(&ret->refcount);
-		local_irq_restore(flags);
-	}
-
 	return ret;
 }
 EXPORT_SYMBOL(get_io_context);
@@ -3574,7 +3530,7 @@ queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 
 	q = container_of(kobj, struct request_queue, kobj);
 	if (!entry->show)
-		return 0;
+		return -EIO;
 
 	return entry->show(q, page);
 }
@@ -3588,7 +3544,7 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr,
 
 	q = container_of(kobj, struct request_queue, kobj);
 	if (!entry->store)
-		return -EINVAL;
+		return -EIO;
 
 	return entry->store(q, page, length);
 }
@@ -3598,7 +3554,7 @@ static struct sysfs_ops queue_sysfs_ops = {
 	.store	= queue_attr_store,
 };
 
-struct kobj_type queue_ktype = {
+static struct kobj_type queue_ktype = {
 	.sysfs_ops	= &queue_sysfs_ops,
 	.default_attrs	= default_attrs,
 };